Method for distinguishing Thangka image from non-Thangka image

The invention discloses a method for distinguishing a Thangka image from a non-Thangka image. The method is implemented on a computer by a training stage and an identification stage, wherein the training stage comprises preprocessing, characteristic extraction, similarity measurement and classifier design, and the identification stage comprises preprocessing, characteristic extraction, similarity measurement and classifier identification. The training stage comprises the following steps of: performing preprocessing operation such as normalization, blocking, graying and the like on an image of a training set; extracting characteristics such as an image information entropy, an image color change rate, image symmetry and the like to obtain identification characteristics for distinguishing the Thangka image from the non-Thangka image; performing template and threshold value training and the similarity measurement on the characteristics of the image information entropy, the image color change rate and the image symmetry; and designing a classifier by adopting an interval threshold incremental algorithm. The identification stage comprises the following steps of: preprocessing the image to be identified; and extracting the characteristics of the image, and classifying and judging whether the image is the Thangka image or the non-Thangka image by using the classifier. The method has high accuracy which reaches 95 percent.